Extinction ratio measurement in optical transmitters.pdf

Measuring Extinction Ratioof Optical TransmittersApplication Note 1550-8

2

Optical transmitters used in high-speed digital communicationsystems are typically required to maintain a specific set of perfor-mance levels. One parameter, extinction ratio, is used to describeoptimal biasing conditions and how efficiently available lasertransmitter power is converted to modulation power. Althoughspecifications are defined by industry standards and test method-ologies loosely described, historically it has been difficult to achieveaccurate and repeatable extinction ratio measurements.

The intent of this application note is to describe the intent of themeasurement, the process used to make an extinction ratio measure-ment, and a methodology to achieve the best possible measurementresults. The note is divided into several sections. (To begin makingmeasurements immediately, proceed to section 4).

1. What is extinction ratio and why measure it, pages 3-5.

2. Extinction ratio measurement processes, pages 6-9.

3. Understanding and maximizing extinction ratio measurement accuracy and repeatability, pages 10-18.

4. The step-by-step procedures for making an extinction ratio measurement using the Agilent 83480A or 86100A Digital Communications Analyzer are found in pages 19-28.

Introduction

3

Extinction ratio, when used to describe the performance of anoptical transmitter used in digital communications, is simply theratio of the energy (power) used to transmit a logic level ‘1’, to theenergy used to transmit a logic level ‘0’. For a graphical description,the eye-diagram is commonly used as shown in Figure 1.

Figure 1. Definition of extinction ratio

Extinction ratio can be determined from the eye-diagram, definedas a linear ratio, in decibels, or as a percentage:

‘1’ power levelExtinction ratio = ———————

’0’ power level

‘1’ power levelExtinction ratio (dB) = 10 log10 ———————

‘0’ power level

‘0’ power levelExtinction ratio % = ——————— x 100

‘1’ power level

Thus if the ‘1’ power level was 1000 microwatts, and the ‘0’ powerlevel was 50 microwatts, the extinction ratio would be 20, 13 dB, or5% depending upon which definition is preferred.

1. What is extinction ratioand why measure it?

‘1’ Power Level

‘0’ Power Level

How does transmitter extinction ratio affect system performance?

The parameter that best describes the overall health of a communicationsystem is bit-error-ratio (BER). Virtually any well designed digitalcommunications system is capable of achieving virtually error-freecommunication if transmitter powers are kept high enough andsystem loss (i.e. fiber attenuation) is kept low enough. To minimizethe need for costly amplifiers or regenerators, it is desirable to havethe longest span possible between the transmitter and receiver.Lengthening the transmission span too far will eventually degradethe system BER, as signal levels drop and noise becomes a dominantcomponent of the signal at the receiver. However, transmitter extinctionratio will also impact the allowable length of a transmission system.Figure 2 shows BER power penalty as a function of extinction ratio.For example, if the extinction ratio is 8.2 dB, approximately 1 dB ofadditional power would need to be transmitted to achieve the sameBER obtained if the extinction ratio were instead 13 dB. In otherwords, a signal with 0 dBm average power and 13 dB extinctionratio should achieve the same BER as a signal with 1 dBm averagepower and 8.2 dB extinction ratio. Similarly, compared to an extinctionratio of 8.2 dB, an additional 1.5 dB of power would be required tomaintain the BER level if extinction ratio were reduced to 5 dB.

4

Figure 2. BER power penalty versus extinction ratio

How does extinction ratio describe transmitter performance?

Good bit error ratio (BER) performance is achieved when there issignificant separation between the power level used to transmit alogical ‘1’ and the power level used to transmit a logical ‘0’. Thedifference between these two power levels describes the modulationpower of the transmitted signal. The larger the modulation power,the easier it will be for the system receiver to accurately determinewhat signal level is present.

Figure 3. Relationships between average power, modulation power, andextinction ratio

3

2

1

0 5 10 15 200

Extinction Ratio (dB)

BER

Pow

er P

enal

ty (d

B)

8.2 13

120011001000

900800700600500400300200100

0

Pmod

Pmod

PAVG

A

B

1000 µW

100 µW

1200 µW

300 µW

PAVG

5

Consider the two eye diagrams of Figure 3. Both signals have thesame modulation power of 900 microwatts (the difference betweenhigh and low levels). Assuming the noise levels for either signal arethe same, the BER’s achieved with either transmitter A or trans-mitter B are likely to be similar. Which condition is superior? Notethat the average power for transmitter A and B are 550 microwattsand 750 microwatts respectively. Transmitter B is required toproduce 200 microwatts additional average power above transmitter Awith no apparent performance improvement.

Ideally, an optical transmitter would be turned off (zero outputpower) in the transmission of a logical ‘0’. However, as a directlymodulated laser approaches or drops below the lasing threshold,system BER performance is degraded due to transmitter wavelengthshift (chirp) and waveform distortion (overshoot and ringing). Thusin Figure 3, neither transmitter A or B is biased and modulated ina manner that causes the laser to operate below the lasing threshold.Transmitter A generates 100 microwatts to transmit a logical ‘0’,300 microwatts for transmitter B. While both conditions then usesome power to transmit a ‘0’, using 300 microwatts to transmit alogical ‘0’ can be considered an inefficient use of available laserpower.

While modulation power described the difference between ‘1’ and ‘0’power levels, it yields no measure of how efficiently laser power isused. However, the ratio of the ‘1’ and ‘0’ power levels, known asextinction ratio, is an indication of efficiency. The extinction ratiofor transmitter A is 1000/100 or 10, whereas the extinction ratio fortransmitter B is 1200/300 or 4. In the limit, extinction ratio canbecome infinite. In this case all the available laser power is convertedto modulation power. (However, this would require the laser to beturned off for the transmission of a ‘0’, and as mentioned above thiswould result in a degraded BER). Extinction Ratio values in therange of 8 to 20 (9 and 13 dB) are common for high-speed, directlymodulated lasers.

For the examples of Figure 3, transmitter A would have an extinctionratio of 10, 10 dB, and 10%. Transmitter B would have an extinctionratio of 4, 6 dB, and 25%.

6

Industry standards have been developed to set minimum require-ments for extinction ratio values and to define a methodology formaking extinction ratio measurements.

ITU G.957: This document defines the physical requirements forSynchronous Digital Hierarchy (SDH) transmitters. Extinctionratio minimums of 8.2 dB to 10 dB are required depending upon theapplication.

Bellcore GR-253-CORE: Defines requirements for SynchronousOptical Network (SONET) systems. Similar to SDH transmitters,extinction ratio minimums of 8.2 dB to 10 dB are required.

IEEE 802.3Z: Defines requirements for Gigabit Ethernet trans-mitters. Extinction ratios greater than 9 dB are specified using testmethods from TIA/EIA-526-4A

ANSI X3.230-1994/X3.297-1997: Defines requirements for FibreChannel transmitters. Extinction ratios greater than 9 dB arespecified using test methods from TIA/EIA OFSTP-4A.

TIA/EIA (Telecommunications Industries of America/Electronics Industries of America) 526-4A (OFSTP-4A):This document describes measurement procedures for digitalcommunications eye-diagrams including extinction ratio.

Extinction Ratio measurements are typically performed on eye-diagrams using digitizing oscilloscopes. Referring to OFSTP-4A,the receiver block diagram for eye-diagram extinction ratiomeasurements is shown in Figure 4. The key elements of the testsystem are an optical receiver and a digitizing oscilloscope. Theoptical receiver consists of a photodiode followed by a fourth-orderBessell-Thomson low-pass filter. The frequency response of thephotodiode/filter combination is precisely specified for both SDHand SONET testing and is shown in Figure 5. This specificationassumes that the oscilloscope does not contribute to the frequencyresponse rolloff.

Figure 4. Test system for measuring extinction ratio

2. Extinction ratiomeasurement process

OpticalInterface

Point

OpticalSignalInput

OpticalTo

ElectricalConverter

Low-PassFilter

Oscilloscope

7

Figure 5. Frequency response requirements for the filtered optical receiver

5

0

–3

–5

–10

–15

–20

–250 0.5 0.75 1.0 1.5 2.0

Normalized Frequency (f / bit rate)

Atte

nuat

ion

(dB)

The frequency response specification provides a consistent measure-ment method for characterizing optical transmitters. For example,the waveform viewed with a very broad bandwidth system is likelyto appear different than when viewed with a reduced bandwidth(filtered) system. See Figure 6. This is a critical issue for eye-diagram mask testing, which is used to define allowable waveformshape. In addition, a reduced-bandwidth (filtered) receiver willbehave similarly to the receivers typically used in an actualtransmission system. That is, system receivers generally have theminimal bandwidth required to accurately operate at a given datarate. Extinction ratio measurements can be made with or without afiltered receiver. In practice, SDH and SONET compliance measure-ments, including extinction ratio are typically made in a filteredbandwidth as the filter performs an integrating effect to simulatethe signal that will be seen by a decision circuit.

8

Figure 6. Waveform measured in a filtered and unfiltered bandwidth

The extinction ratio measurement is made through a statisticalanalysis of the eye-diagram. OFSTP-4A recommends that the meanvalue of the logical ‘1’ signal and the mean value of the logical ‘0’signal be determined at the center of the eye. Again, the ratio ofthese two values yields the extinction ratio.

9

Figure 7. Using histograms to determine extinction ratio

A histogram analysis is used to determine the ‘1’ and ‘0’ levels.Figure 7 shows an eye-diagram with a histogram constructed alongthe vertical axis for the central 20% of the eye-diagram. As expected,the histogram is bimodal. The majority of the time the waveformexists at either the logic 1 level or the logic 0 level. To determinethe mean values discussed above, the histogram is split in halfvertically. The mean value of each of the two histograms is thendetermined. Ideally this should be sufficient information from whichto determine the extinction ratio. But as is the case with manymeasurements, there are factors which can lead to measurementuncertainty and inaccuracy.

10

To minimize measurement degradation, the following strategyshould be implemented:

• Use a histogram-based measurement algorithm that is robust even when error producing situations are present

• Understand the causes of waveform distortion and use instrumen-tation that yields high waveform fidelity

• Understand and where possible stay within the measurement limitations of the instrument

Offsets

It is a common occurrence for photodiode receivers to generate anon-zero output voltage when no light is present at the input. Thiscan occur due to photodiode dark currents, or can be generated byelectrical amplifiers following the photodiode. Also, offsets can begenerated by the oscilloscope following the photodiode. Consider thesimple eye-diagram of Figure 8. With a ‘1’ level of 1 milliwatt and a‘0’ level of 50 microwatts, the extinction ratio is 20, 13 dB, or 5%. Ifthis signal is passed through a receiver with an optical to electricalconversion gain of 500 Volts per Watt and a ‘dark’ or offset voltageof 10 millivolts, how will the waveform on an oscilloscope appear?

Figure 8. How offsets can affect measurement accuracy

At the receiver output the ‘1’ level will be shifted from an idealvalue of 500 millivolts to 510 millivolts. The ‘0’ level will be shiftedfrom an ideal value of 25 millivolts to 35 millivolts. The apparentextinction ratio would be 14.6, 11.6 dB, or 6.9%. With the offsetunaccounted for, the measurement error would be:

measurement error = ((20-14.6)/20)*100 = 27%

1000 µW

50 µW

510 mV

35 mV

O/E

3. Understanding andmaximizing extinction ratiomeasurement accuracyand repeatability

There are several factors which can potentially degrade an extinctionratio measurement. These factors can be grouped into the followingcategories:

• Offset/spurious signals generated by the instrumentation• Distortion of the waveform caused by the instrumentation• Precision of the instrument in measuring the amplitudes of the

waveform

11

When measuring transmitters with higher extinction ratios, aneven larger measurement error will be experienced as the offseterror approaches and may even dominate the true ‘0’ level. In thisspecific example, without removing the offset error, the highestextinction ratio that can be measured (for a constant 500 millivolt‘1’ level) would be 510/10 or 51 (17 dB, 2%), even if the true extinctionratio was infinite. Note also that a negative offset will potentiallycause an extinction ratio measurement result to appear to be muchlarger than the true value.

Another issue arising from offsets occurs if the offset fluctuates.The result is an extinction ratio measurement that has variation aswell as inaccuracy.

Reducing measurement error due to instrumentation offsets

Both the HPAgilent 83480A and Agilent 86100A Digital Communi-cations Analyzers provide simple and straightforward ways todecrease extinction ratio measurement error due to offsets. In theAgilent 83480A this process is referred to as dark calibration or“dark cal”. In the Agilent 86100A the process is referred to as anextinction ratio calibration. In either instrument the procedure isessentially the same. The instrument requests that any source oflight be blocked from entering the optical receiver. The instrumentwill then measure any residual signals present when there is noinput to the receiver. When an extinction ratio measurement isperformed, the instrument will mathematically remove the offsetfrom the extinction ratio calculation.

Both the Agilent 83480A and 86100A have integrated opticalreceivers which are designed to minimize and stabilize internaloffset signals. If external optical receivers are used, the calibrationalgorithms are still valid when performed on an electrical channelto which the receiver is attached. It is important to ensure that theexternal receiver be attached and active when the calibration isperformed. In this way any offsets generated by the receiver will beaccurately characterized.

Effects of instrumentation frequency response

The frequency response of the measurement system, including thephotodiode O/E converter, any amplification, filtering, and themeasuring oscilloscope can potentially lead to waveform distortionand an eventual degradation of extinction ratio measurements.

TIA/EIA OFSTP-4A recommends that the frequency response of thereceiver, including filtering, follows the response window previouslyshown in Figure 5. The allowable tolerances are very tight and canbe difficult to achieve in practice. However, it is possible that areceiver approaching these tolerances can have a frequency responsethat results in an extinction ratio measurement with significantinaccuracy.

12

Figure 9. Receiver frequency response causing extinction ratio measurementinaccuracy

Consider a receiver with a frequency response similar to thatshown in Figure 9. It almost falls within the required specificationsfor a SDH/SONET reference receiver and might then be used forextinction ratio compliance measurements. Note that the frequencyresponse appears well behaved, except at the very low frequencyrange, where a distinct uplift occurs approaching 1 dB. Theimplication of this response is that very low frequency signalcomponents will experience amplification relative to the middle andhigh frequency components of the data signal. Increased lowfrequency gain is a common phenomenon of GaAs amplifiers usedin high-speed, wide-bandwidth O/E receivers or off-axis coupling oflight to the photodiode.

When a waveform such as the signal shown in Figure 10 is passedthrough a measurement system with the frequency response similar toFigure 9, it will experience distortion as the low frequency compon-ents will be amplified relative to the rest of the signal. A trans-mitter with an approximate 10 dB extinction ratio is measured at5.48 dB. The easiest way to view the impact this has on an extinctionratio measurement is to classify the low frequency components assimply the DC or average power and the high frequency componentsas the AC modulation or information bearing element of the signal.

5

0

–5

–10

–15

–200 5•10 1•10 1.5•10 2•10 2.5•10 3•10 3.5•10 4•10

Frequency (Hz)

Mag

nitu

de (d

Be)

8 9 9 9 9 9 9 9

13

Figure 10. Waveform distortion due to imperfect frequency response

The effect of this offset is similar to that caused by dark levels (seeFigure 8). As true extinction ratios become high, the measurementerror can increase significantly. However, unlike dark levels, thereis no simple technique to remove offsets due to the frequencyresponse of the receiver. This is because the correction required isdependent upon the frequency content specific to the signal beingmeasured.

The best technique for minimizing extinction ratio measurementerrors due to frequency response imperfections is to use a measure-ment system designed to minimize this effect*. Caution should beused when measurements are made with amplified receivers.Receivers that are not amplified, or that employ carefully designedamplification are preferred for accurate extinction ratiomeasurements.

Receivers used for making extinction ratio measurements includingthe Agilent 83481A, 83482A, 83485A, 83485B, 83486A, 83487A,86101A, 86103A, 86105A, 86106A, and 86109A are designed to havewell behaved frequency responses that yield very accurate extinctionratio measurements. An example of the frequency response of theAgilent 83485A is shown in Figure 11.

* In limited cases, frequency response error at a single bit rate can be removed using the techniquedescribed in Appendix 1)

14

Figure 11. Frequency response of the Agilent 83485 Optical receiver

Extinction Ratio measurement algorithms

The process used by the Agilent 83480A for automatically measuringextinction ratio consists of the following elements:

1. Remove internal offset effects through a dark calibration

2. Set up the instrument in an infinite persistence database mode (This is a 3 dimensional database of number of samples vs. time and amplitude. This is displayed in a color graded mode where colorrepresents the number of samples measured at a given display pixel.)

3. Perform a histogram analysis of the eye-diagram to determine the ‘1’ and ‘0’ levels of the eye diagram

4. Calculate the extinction ratio and report the results

The algorithms used for making extinction measurements haveevolved over time. When the Agilent 83480 was introduced in 1994,there was not a strong concensus within the industry on what regionof the eye diagram should be used for analysis. Some felt the measure-ment should be made over the full bit period of the eye. Others feltthat only data from the central 20% of the eye should be used. Tomake the instrument as flexible as possible, it was designed to beconfugurable to measure any region of the eye the user desired.

When wide regions are analyzed, the rising and falling edges of theeye may potentially influence the measurement even though theseportions of the waveform are not typically considered part of theactual ‘1’ or ‘0’ levels. If data from the rising and falling edges of theeye diagram are included in the histogram, they will tend to raisethe apparent mean ‘0’ level and lower the apparent ‘1’ level. Thusearly algorithms were designed to isolate data from the rising andfalling edges.

10

–1–2–3–4–5–6–7–8–9

–10–11–12–13–14–15

10.0 347.3 684.5 1021.8 1359.1 1696.4 2033.6 2370.9 2708.2 3045.5 3382.7 3720.0

Frequency (MHz)

Res

pons

e (d

Be)

CCITTWindow

15

This feature of the extinction ratio measurement algorithm wouldsometimes cause problems with eye diagrams that exhibited patterndependency. For example the ‘1’ levels following another logic ‘1’might be consistently higher than ‘1’ levels preceded by a logic’0’.The eye diagram would thn exhibit 2 discrete ‘1’ levels. The histo-gram for the ‘1’ level would then exhibit two distinct modes. To theanayzer, this has a similar signature to the histogram for the ‘1’level when using data from the full bit period of an eye due to datafrom rising and falling eges. The analyzer would then sometimes“lock” onto one of the discrete modes of the bimodal histogram whenin reality it should have found the overall mean of the entire dataset. This might occur even if only the central region of the eye wasbeing analyzed. If neither of the modes were dominant, the analyzermight momentarily lock onto one mode and then lock onto theother. The effect would be seen in the reported value of extinctionratio. It would bounce back and forth between two values. (If theinstrument markers were tracking the measurement, they wouldalso jump in position.)

Since 1994, the industry has generally concluded that an extinctionratio measurement should be made over the central region of theeye diagram. Because of this, the measurement algorithm forextinction ratio has been optimized for this case. The new algorithmwill simply find the overall mean of the data within the measure-ment boundaries (typically the central 20% of the eye). It does nottry to interpret whether any of the data is from a rising or fallingedge. There are two distinct benefits from this change. First, eyediagrams with pattern dependency no longer suffer from measure-ment “bounce”. Second, measurement results are achieved muchquicker and with significantly better repeatability and stability.The updated measurement algorithm has been implemented infirmware revision 7.0 for the Agilent 83480A and for all revisions of the Agilent 86100.

Regardless of the measurement algorithm, the following issues areimportant to consider:

• Will noise degrade the measurement?• What is the smallest signal level that can be measured accurately?• What portion of the eye is analyzed?• What constitutes an adequate amount of data from which to

construct histograms?

Instrumentation noise and measurements of small signals

Instrumentation noise can also lead to extinction ratio measure-ment error. However, because the measurement is histogram based,it can be very robust even as noise levels approach the magnitudeof the signal being measured. For example, the RMS noise level ofthe Agilent 83480A when using the Agilent 83485A optical receiverplug-in or Agilent 86100A with the 86105A plug-in is typically 8 microwatts (–21 dBm). This noise level will be seen on both the ‘1’and ‘0’ levels for a combined RMS level of 16 microwatts.

16

Transmitter signals at average power levels of lower than 40 micro-watts (–14 dBm) at an extinction ratio on the order of 10 can still beaccurately measured even though the eye pattern begins to close.

Figure 12. Extinction Ratio measurements with low signal-to-noise ratios

Using the automatic histogram capabilities of the instrument, themean ‘1’ and ‘0’ levels from which extinction ratio is calculated areeasily determined. In the case of Figure 12, the eye height (differencebetween ‘1’ and ‘0’ levels)-to-RMS-noise ratio approaches 10, wherethe RMS noise is 16 microwatts (double 8 microwatts) and thesignal power is approximately 100 microwatts (roughly double theaverage power when the extinction ratio is greater than 10). As this ratio goes much below 5, the extinction ratio measurementalgorithm will eventually fail as the instrument can no longerdifferentiate the components of the eye diagram as required to setup the location of the histograms.

There is a large family of receiver plug-ins compatible with the Agilent 83480A and 86100A, each with their own noise characteristic.The Agilent 83481A, 83486A, 83487A, 86101A, and 86103A exhibittypical RMS noise levels of less than 1.5 microwatts, allowingextinction ratio measurements on signals as low as 10 microwatts(–20 dBm avg.). The Agilent 83485B and 86106A RMS noise levelsare typically less than 15 microwatts, allowing extinction ratiomeasurements on signals as low as 100 microwatts (–10 dBm).Although extinction ratio measurements can be made with lowSNR’s, it should be noted that mask testing of eye-diagrams requirean SNR on the order of 12 or greater. For example, when using theAgilent 83480A and 83485A (RMS noise of 8 microwatts on both ‘1’and ‘0’ levels) the signal power should exceed 200 microwatts(average power in excess of 100 microwatts, –10 dBm) for mask testing.

17

Determining what portion of the eye to measure

OFSTP-4 recommended that the histogram from which the extinctionratio data is extracted include data from a full bit period of the signal orin other words one complete “eye”. An advantage of this methodologyis that it allows for simplified alignment of the histogram window.As long as the horizontal width of the window is one full bit period,the horizontal positioning of the window is trivial. The Agilent83480A and 86100A allow the histogram to be taken from any portionof a bit period. For example, the default setting of the Agilent 83480Aand 86100A is to measure the central 20% of the eye.

The central 20% of the eye is used for two reasons. First, the mostrecent revision of OFSTP-4A specifies that extinction ratio measure-ments be made here. Second, this is the portion of the filtered eyethat represents the true mean ‘1’ and ‘0’ powers of the unfilteredeye waveform. If measurements are to be made on a portion of theeye other than the central 20%, this parameter can be modifiedin the Agilent 83480A under the DEFINE MEAS, Color Grade, Eyewindow menu. In the Agilent 86100A this parameter can be alteredunder the Measure, Config Measure, Eye Boundary setting.

Acquiring an adequate database

Measurements based upon statistical sampling and histogramanalysis are generally improved as the sample size is increased.with firmware revisions earlier than 7.0 for the Agilent 83480A,measurements would tend to fluctuate and then eventually stabilizeas the number of waveforms displayed in infinite persistence (colorgrade mode) increased.

Figure 13 shows the general shape of the vertical histogram for thecase where a single waveform has been acquired versus the casewhere 50 waveforms have been acquired. Note that with a singlewaveform the histogram is rather coarse, and with 50 waveformsthe histogram is very smooth indicating a well-filled sample space.

The 83480 (firmware revision 7.0 and higher) and 86100 aredesigned not to report an extinction ratio result until at least onepixel on the display has been hit at least 15 times. 20 or morewaveforms may need to be aquired to reach this density of data.The 83480A allows this parameter to be altered through DEFINEMEAS, Color Grade, # Hits. The 86100 allows this to be alteredonly through remote command.

As the extinction ratio measurement algorithms have been improved,accurate and stable measurement results can be achieved for just afew waveforms. This can be examined by reducing the # Hitsparameter and observing the stability of the reported meassrementas just a few waveforms are initially collected versus a large numberof waveforms. In general, histograms from even small sample sizesyield an accurate assessment of the mean ‘1’ and ‘0’ levels.

18

Figure 13. Extinction ratio histograms for small and large sample spaces.

19

The following section discusses the step-by-step procedure formaking an extinction ratio measurement using the Agilent83480A and 86100A Digital Communications Analyzers. Followingthese guidelines will provide optimum measurement results.

Start with a known instrument configuration

To begin the procedure from a known instrument state, the instru-ment can be placed into a default setup condition by pressingSETUP (83480 only) and the Default Setup softkey. Note that thiswill revert any customized instrument configurations back to theirdefault conditions. While the default setup is not required toperform an extinction ratio measurement, it does guarantee thatunusual instrument configurations will not impact the results.

Getting the signal on the instrument display

The Agilent 83480A and 86100A, like all very wide-bandwidth“equivalent-time” digitizing sampling oscilloscopes, requires anexternal trigger for a timing reference. The best trigger signal is aclock signal that is synchronized with the data. For example, if a2.5 Gbit/s waveform is to be measured, the best trigger signal is the2.5 GHz clock signal used to generate the signal to be measured.This timing signal could come from the pattern generator used toproduce the data that modulated the laser, or a clock signal extractedfrom the data. Subrate clock signals may also be used if they areinteger divisions of the actual data rate (i.e. a divide by 4 or divideby 10 clock, 622.08 or 248.832 MHz for a 2488.32 Mbit/s waveform).The trigger signal must be phase-locked or phase coherent to thedata waveform. Since all measurements are made relative to thetrigger event, any relative phase instability between the triggersignal and the data will be seen as waveform instability on theinstrument display.

In general, extinction-ratio measurements are not sensitive to mildtrigger vs. data instability (jitter) in that the measurements aremade on signal amplitudes. A mask test on the other hand can beseverely impacted by jitter.

When measuring optical waveforms, a high-integrity connectionshould be made to the optical receiver. Poor connections can causeamplitude fluctuation and degrade measurements. Make sure thatthe fiber ferrule of both the instrument receiver and signal fiber areclean and in good mechanical condition. Once the connection ismade, the instrument can be configured for the extinction ratiomeasurement.

4. Procedure for measuringextinction ratio using theAgilent 83480A or 86100ADigital CommunicationsAnalyzer

20

Extinction ratio measurements require that only a single channelbe active. Turn off the other channels by selecting the ChannelSETUP key (found above the input connectors on Agilent 8348Xplug-in modules) and turning the channel OFF for each channelbesides the one to be measured. With the Agilent 86100A, extinctionratio measurements are made while operating in the “EYE/MASK”mode. Pressing the Eye/Mask button to the right of the instrumentdisplay activates this mode. When this is selected, the instrumentautomatically will turn off all input channels but one and put theinstrument in an infinite persistence mode.

Filtered or unfiltered?Extinction ratio measurements can be made in either a filtered orunfiltered bandwidth (although unfiltered measurements do notcomply with the OFSTP-4A test method). In the Agilent 83480A,the filter configuration can be setup by pressing the ChannelSETUP key, Bandwidth/Wavelength (softkey), Filter (on or off). Inthe Agilent 86100A, the filter can be activated by selecting theSetup dialog and selecting the appropriate channel, or by pressingthe approriate channel button below the display graticule. Someplug-in modules have dual filters for two data rates. If the filter ison, make sure the dorrect filter has been selected.

Setting the vertical and horizontal scalesThe measurement should be made with a complete eye-diagram onthe instrument display. Thus the vertical and horizontal scales shouldbe adjusted to achieve this. Using the Agilent 86100A, scaling is easilyachieved by simply pressing the autoscale key. Although the Agilent83480A has an autoscale function, it is not designed to optimallydisplay an eye-diagram. The following can be used to display the eye.

Press the CHANNEL SETUP key on the plug-in for the channel tobe measured. Press the Channel Autoscale softkey. This will do avertical scaling without altering the timebase settings. Press theTimebase key. Set the time base [Time/Bit period] key to bit period.Set the Bit Rate key to match the data rate of the signal (such as2488 Mb/s STM-16/OC-48) either by using the knob or arrow keysto scroll through the built-in standard rates, or enter the data ratethrough the numeric keypad. Set the Bit Period to a value betweenone and two bits. Then adjust the Position value to center the eye-diagram on the screen.

Activating the color graded databaseExtinction ratio measurements are made using a histogram analysis ofthe eye-diagram. Using the 86100A the database for histograms issimply any data on the infinite persistence display that isautomatically activated when the instrument is placed in Eye/maskmode. For the 83480A, the database from which the histogram isderived is the color graded database. Thus the instrument must beplaced in the color graded display mode. This is accomplished bypressing the DISPLAY key and then the Color Grade softkey(turning it “on”). As soon as Color Grade is activated, the databasewill begin to accumulate data. However, for both the 83480A and86100A when any of a variety of settings are altered on theinstrument (examples include vertical or horizontal scaling), thedatabase will be refreshed and begin to build again.

21

Setting up the extinction ratio measurement

Agilent 83480AThe extinction ratio measurement is found under the Meas Eyemenu (located as a “shift” function of the Acquisition key below theinstrument display). When extinction ratio is selected, the extinctionratio menu will appear. First select the format for reporting results,either linear ratio, decibels, or as a percentage.

Agilent 86100AThe extinction ratio measurement is available as one of the automaticeye measurements found on the left side of the display when theEye/Mask mode is activated (key found to the right of the display).With an eye diagram on the display, the extinction ratio is reportedafter pressing the “extinction ratio” measurement key. In its defaultstate, the 86100A reports extinction ratio in decibels. To change tolinear or percentage results, press “setup and info” on the measure-ment results rab (to the right of the reported extinction ratio value).Then press “extinction ratio” and “configuration”. The extinctionratio format can then be changed.

Figure 14a. Extinction ratio menu for the 83480A

22

Figure 14b. Extinction ratio menu for the 86100A

An extinction ratio or “dark” calibration should then be performedto allow internal offsets (either of an integrated optical channel oran external O/E converter and electrical channel) to be removedfrom the measurement. However, prior to performing the dark calsome discussion on vertical scale and offset is required. The Agilent83480A and 86100A are a digitizing oscilloscopes. Measured signallevels are converted to digitized levels. To obtain the highest accuracyin the digitizing process, the vertical scale and offset should beadjusted so that the eye-diagram covers most of the instrument display.On the other hand, when measuring the dark signal level (thesignal level of the instrument when no signal is applied) during anextinction ratio or dark cal, the vertical scale and offset must besuch that the dark level is also visible on the instrument display.

To satisfy both of these conditions, first adjust the vertical scale and offset to present the largest eye possible while leaving about halfof a vertical graticule between the bottom of the eye and the bottomof the instrument display. Specific vertical scale values such as 80 microwatts/div will likely need to be entered through the numerickeypad instead of the 50/100/200....settings achieved through thearrow keys or knob. Disable the light source to be measured byeither turning it off or disconnecting it from the instrument. Verifythat the dark level is on the screen. It should appear as a flathorizontal line, and should be located near the bottom of the screen.If it is not on screen, or is perhaps difficult to see due to the displaypersistence still showing the eye-diagram in the ‘on’ condition, thevertical scale or offset should be adjusted until the dark level isvisible. The vertical scale adjustment process can then be repeateduntil the largest eye diagram is visible while still being able to displaythe dark level when the light source is disabled. See Figure 15.

23

Figure 15. Optimum vertical scale displays the dark level and the largesteye possible.

Setting the vertical scale to maximize the size of the eye-diagram isnot a rigorous requirement for an accurate extinction ratio measure-ment. It is a recommendation for the best possible measurementaccuracy. Valid, accurate results will still be obtained for eye-diagramswith as little as 3 graticules of vertical display magnitude.

Performing an extinction ratio or “dark” calibration

Extinction ratio calibration for the Agilent 86100AThe Agilent 86100A will report an extinction ratio value, even withouta valid extinction ratio calibration. However, for best accuracy anextinction ratio calibration should be performed. Once an extinctionratio calibration has been performed, it will remain valid as long asthe temperature of the instrument does not vary more than ±1°C orless than 10 hours has elapsed since the calibration was performed.Thus the calibration should be performed after the instrument haswarmed up, typically achieved after one hour. Also, the calibrationshould be perfomed at the vertical scale setting at which themeasurement will be made. If the scale setting is changed morethan one major scale setting, a new extinction ratio calibration isrecommended. (For example, if the scale was 200 µW/division whenthe calibration was performed, this calibration is adequate for scalesettings from 100 µW/division to 500 µW/division).

To perform an extinction ratio calibration, select the calibration pulldown menu at the top of the display. This will then display a pagethat icludes all the available calibrations for the instrument. Selectthe tab for extinction ratio. The display should appear as Figure 16a.

Dark level (typically not visible simultaneously with the eye diagram)

24

Figure 16a. Calibration status of the plug-in channel in the 86100A

The extinction ratio calibration page will indicate whether a calibrationis recommended, why it is recommended, and the conditions underwhich any current calibrations have been performed. To execute acalibration, select the “Ch X Extinction Ratio Calibration” button.

When the calibration is selected, the user is instructed to removeany signals present at the input to the instrument. It is best toblock any light going into the receiver, as stray ambient light cancorrupt a calibration, particularly those with large core diameters.If external optical receivers are used, the receiver should be attachedto the electrical channel of the instrument and be in an active state.

If the plug-in module does not have a valid vertical calibration, theextinction ratio calibration will automatically invoke a verticalcalibration. This will take one to two minutes. Once the verticalcalibration is performed, or a valid vertical calibration alreadyexisted, the extiction ratio calibration is performed. This takes lessthan two seconds. (This is the reason that sometimes extinctionratio calibrations take one to two minutes or just one or two seconds.)

Extinction ratio calibration for the Agilent 83480AA dark calibration is required to allow the instrument to removeany internal offset signals from the extinction ratio calculation. If avalid dark calibration has not been performed, the instrument willnot report any extinction ratio results. If a dark calibration hasbeen performed and the instrument is turned off, the dark calibrationis invalidated. A similar condition will occur if the ambienttemperature of the instrument changes more than 5 degrees C.

25

A dark calibration is performed under the MEAS EYE/ExtinctionRatio/Dark Cal key sequence. When Dark Cal is pressed, the instru-ment instructs the user to remove all signal connections to the plug-in.The intent is to have no optical signal power going into the instru-ment so that what is measured is only due to internal offsets. In thecase where an external receiver is used, offsets due to both thisreceiver and the Agilent 83480A are measured. When using receiverswith large core diameters such as the Agilent 83486A/7A, ambientlight can be significant enough to corrupt the calibration. Thereforethe input to the receiver should be blocked so no light can enter.

For best accuracy, the dark cal should be performed after theinstrument has warmed up. A one hour warm-up time is sufficient.A plug-in calibration will also ensure best accuracy. This is easy toperform and is achieved under the UTILITY/Calibrate/CalibratePlug-in key sequence. (A plug-in cal should be performed wheneverthe ambient temperature of the instrument has deviated more than5 degrees C from the temperature value when the last plug-in caltook place, or the plug-in module has been removed from the main-frame. The temperature status can be seen by selecting SETUPCHANNEL near the measurement channel input and selecting theCalibrate/Cal Status softkeys and turning the Cal Status “ON”. SeeFigure 16b. Note that in this example the temperature deviationfrom when the plug-in calibration (not the mainframe calibration)was performed is 0 degrees C. Once the plug-in cal has been performed,a dark cal will still be required.)

Figure 16b. Calibration status of the plug-in module in the 83480A

26

When the dark cal is first performed with an integrated opticalchannel, two procedures take place. An O/E offset calibration isexecuted. In this process a partial vertical calibration for the plug-in module takes place which will minimize any internal offsets fromthe internal optical receiver and the following electronics. Second,the actual dark level after offset adjustment is measured. Thisprocedure takes approximately 35 seconds. If the dark cal procedureis performed a second time, the O/E offset calibration is not requiredand only a dark level measurement is performed. In this case, thedark cal process takes less than 2 seconds. Note that for bestaccuracy, the dark cal should be performed with the same verticalscale and offset values that will be used when the extinction ratiomeasurement is made. If signal levels change (example, a secondtransmitter with lower power is tested), the dark cal should berepeated once the vertical scaling has been adjusted appropriate forthe new laser under test. If a dark calibration is not repeated, themeasurement results are still valid and accurate. Repeating thedark calibration simply guarantees the best possible accuracy.

Once a dark cal has been executed, the instrument instructs theuser to reconnect the laser. Because the instrument has likelyacquired data before the laser is reconnected, the database shouldbe refreshed to ensure that it contains data only from the activelaser signal. This is achieved by pressing CLEAR DISPLAY oncethe laser signal has stabilized. To complete the procedure, pressDONE in the Extinction Ratio menu and the measurement resultshould then be reported.

General issues with extinction ratio calibrations

It is important to note that the calibration does not remove or “zeroout” any offset signals. It only quantifies them and allows for themto be mathematically removed from the measurement calculation.

As discussed in an earlier section (page 17), the Agilent 83480A allowsthe user to set the minimum number of “hits” or samples that mustbe acquired before a valid extinction ratio measurement is calculated.For example, if the minimum number of hits is set to 15, at leastone of the display pixels must be hit at least 15 times before theextinction ratio is calculated and displayed. This forces the data-base to build up to a significant sample size prior to reporting results.This limit is user defined and can be adjusted under DEFINEMEAS/ Color grade/ # Hits. The default value is 1, meaning thatmeasurement results are yielded immediately (upon display of thefirst color grade waveform). If the # Hit limit is set to a small valuethere will possibly be fluctuation in the measurement results as thedatabase will initially be small. However, as data is continuouslyacquired, the extinction ratio measurement should quickly convergeto a final value.

27

In the Agilent 86100A, controlling the #hits is currently onlyachievable through GPIB control. The default value is 15 hits.Depending upon the signal distribution of typical signals, it islikely that stable, repeatable measurements can be achieved with a#hits setting much lower than 15. The result is that fewer waveformsand less time is required to perform the measurement.

If a laser under test is being actively tuned, then the database willneed to be periodically refreshed so that data taken under previoustuning levels/old operating conditions does not impact the currentmeasurement. Refreshing the database is achieved through executinga CLEAR DISPLAY.

How accurate are the results?

The accuracy of the extinction ratio measurement made by the Agilent 83480A is dependent upon the actual level of extinctionratio. As extinction ratio levels become high, eventually theinstruments ability to make an accurate measurement is limited byits dynamic range. In these cases the instrument is required tosimultaneously measure a very large level and a very small level.The vertical accuracy for this type of measurement is described bythe “DC accuracy” or the accuracy in making a measurement of aspecific signal level. For example, using the Agilent 83485A plug-in,the DC accuracy is ±25 µW ±2% of the measured level. If the logic‘1’ level is 1000 µW, and the ‘0’ level is 100 µW, the measurementextremes (ignoring offsets) would be:

(1000*1.02 + 25)/(100*0.98 –25) = 14.32 (11.6 dB) to

(1000*0.98 –25)/(100*1.02 + 25) = 7.52 (8.8 dB)

If the same analysis is used on a 16 dB (40:1) extinction ratio, themeasurement uncertainty for a 1000 µW ‘1’ level and a 25 µW ‘0’level becomes:

(1000*1.02 + 25)/(25*0.98 –25) = (Infinite) to

(1000*0.98 –25)/(25*1.02 + 25) = 18.9 (12.8 dB)

The above analysis ignored offsets. However, the precision withwhich offsets are measured will also impact the final result. Tocompute extinction ratio, the internal offset determined throughthe dark cal is subtracted from both the logic ‘1’ and logic ‘0’ levelsof the eye diagram. When the logic ‘0’ level is very small (as is thecase when extinction ratio is high), removing the offset requires thesubtraction of a small number from another small number. Again,because both the logic ‘0’ level and dark level are measured with avertical scale setting that also allows the logic ‘1’ level to be measured,the digitization process for the small signals has a comparativelylow level of precision (subject to quantization error). Thus the valuethat dominates the extinction ratio measurement, the offset adjustedlogic ‘0’, can be subject to significant measurement uncertainty.

28

The above analysis would tend to cause a general distrust of extinc-tion ratio measurements. In reality, measurements are significantlybetter than that implied by a worst case uncertainty analysis. Thesituation is alleviated in large part through the use of histograms.Histograms tend to smooth out some of the effects of quantizationerror and yield an effective improvement in measurement accuracy.To give a practical assessment for measurement accuracy, thefollowing tables show “measured versus actual” performance for avariety of Agilent 83480A and 86100A configurations.

Table 1. Agilent 83480A Extinction Ratio Measurement Accuracy

Actual Extinction Ratio (dB)

8 to 11 11 to 14 14 to 16

Plug-in Module Measurement error

Agilent 83485AAgilent 86105AAgilent 86106A –0.6 to 0 –1.5 to 0 –2 to 0

Agilent 83481AAgilent 83486AAgilent 83487A –0.4 to 0 –1.2 to 0 –2 to 0Agilent 86101AAgilent 86103A

Notes: Measurements based upon a color graded database with 50 waveforms, 2^7-1 PRBS data. Measurements madeover the central 20% of the eye-diagram. Eye-diagrams typically scaled to approximately 6 or more vertical divisions.Horizontal scale set to show approximately 1.3 bit periods. Measurements made in a filtered bandwidth (SDH/SONET orGigabit Ethernet Filtering according to the data rate). Valid plug-in vertical calibration and extinction ratio dark calibration.Average power levels from –4 dBm to –14 dBm for the Agilent 83485A plug-in module. Average power levels from –10 dBmto –18 dBm for the Agilent 83481A, 83486A and 83487A plug-in modules. Measurement uncertainties may be different forwaveforms with unusual waveshapes leading to poorly shaped vertical histograms. These results are not instrumentspecifications, but are intended to give expected measurement performance levels.

What levels of extinction ratio can be measured?

Although in theory extinction ratio can be infinite, the measurementof very high extinction ratios is difficult to achieve with any precisiondue to the measurement uncertainties discussed above. Dependingupon the data rate and instrument configuration, Table 1 shouldgive an indication of both measurement range and accuracy.

A technique developed to extend extinction ratio measurementrange has been implemented in the Agilent 83480A and is discussedin Appendix 1. This technique is currently not available in theAgilent 86100A.

29

If a transmitter with an adjustable extinction ratio was measuredby the Agilent 83480A, and transmitter adjustment produced ahigher and higher extinction ratio, eventually the extinction ratioreported would reach a maximum even if the true extinction ratiocontinued to increase. If this measurement limit can be determined,the Agilent 83480A will allow a user-entered adjustment to increasethe useable measurement range. A laser source with an adjustable,very high extinction ratio is required. The procedure is as follows:

1. Perform all the steps required for an extinction ratio measure-ment as described in this application note. When selecting the format for the measurement result, select % as opposed to linear or dB.

2. Beginning with an extinction ratio in the range of 10%, system-atically adjust the extinction ratio for a lower and lower % level. (Recall that as extinction ratio goes up, extinction ratio percentagewill go down). After each laser adjustment, execute a CLEAR DISPLAY to refresh the database.

3. At some point, the reported extinction ratio percentage will no longer change. Either the instrument has reached a measurementlimitation or the laser has reached maximum adjustment. To determine which is the case, one tool that can be used for a directly modulated laser is to monitor the average power (Shift/ More Meas (on the numeric keypad)/Avg Power). If the average power continues to change but extinction ratio does not, it is likely that the instrument is at its limit. The waveform could also be observed to see if it continues to change with adjustment.

4. When the measurement limit for the instrument has been deter-mined, document this value. This may be somewhere between 1 and 4% and will depend upon the instrument configuration and the data rate being measured.

5. In the extinction ratio menu (MEAS EYE/Extinction ratio) there is a key titled “Ch ‘x’ freq corr”. Select this key and through the numeric keypad enter the percentage value obtained in step 4. For example, if the measurement limit is 1.5%, press 1, ‘.’, 5 and enter on the keypad.

Appendix 1: Extendingextinction ratio measurementrange

30

Figure 17. Frequency correction menu

Subsequent extinction ratio measurements will be offset to accountfor the measurement floor of the instrument. The format can bereturned to linear or dB if desired and still use this correction.Note that this correction factor is only valid for the conditions forwhich it was determined. That is, the frequency content of the datashould be constant. If a PRBS data sequence is used, the patternlength and data rate must be held constant. For a detailed explan-ation of the theory behind this procedure, refer to “Accurate OpticalExtinction Ratio Measurements”, IEEE Photonics Technology Letter,Vol. 6, No. 11, November 1994, P.O. Andersson and K. Akermark.

Agilent Technologies’Test and Measurement Support, Services, and AssistanceAgilent Technologies aims to maximize the value you receive, while minimizing your risk and problems. We strive to ensure that you get the test andmeasurement capabilities you paid for and obtain the support you need. Our extensive support resources and services can help you choose the rightAgilent products for your applications and apply them successfully. Every instrument and system we sell has a global warranty. Support is availablefor at least five years beyond the production life of the product. Two concepts underlie Agilent’s overall support policy: “Our Promise” and “YourAdvantage.”

Our PromiseOur Promise means your Agilent test and measurement equipment will meet its advertised performance and functionality. When you are choosing newequipment, we will help you with product information, including realistic performance specifications and practical recommendations fromexperienced test engineers. When you use Agilent equipment, we can verify that it works properly, help with product operation, and provide basicmeasurement assistance for the use of specified capabilities, at no extra cost upon request. Many self-help tools are available.

Your AdvantageYour Advantage means that Agilent offers a wide range of additional expert test and measurement services, which you can purchase according to yourunique technical and business needs. Solve problems efficiently and gain a competitive edge by contracting with us for calibration, extra-cost upgrades,out-of-warranty repairs, and on-site education and training, as well as design, system integration, project management, and other professionalengineering services. Experienced Agilent engineers and technicians worldwide can help you maximize your productivity, optimize the return oninvestment of your Agilent instruments and systems, and obtain dependable measurement accuracy for the life of those products.

By internet, phone, or fax, get assistance with all your test & measurement needs.

Online assistance:www.agilent.com/comms/lightwave

Phone or FaxUnited States:(tel) 1 800 452 4844

Canada:(tel) 1 877 894 4414(fax) (905) 282 6495

Europe:(tel) (31 20) 547 2323(fax) (31 20) 547 2390

Japan:(tel) (81) 426 56 7832(fax) (81) 426 56 7840

Latin America:(tel) (305) 269 7500(fax) (305) 269 7599

Australia:(tel) 1 800 629 485 (fax) (61 3) 9210 5947

New Zealand:(tel) 0 800 738 378 (fax) 64 4 495 8950

Asia Pacific:(tel) (852) 3197 7777(fax) (852) 2506 9284

Product specifications and descriptions in this document subject to change without notice.

Copyright © 1998, 2000 Agilent TechnologiesPrinted in USA January 1, 20015966-4316E